Update app.py

app.py

@@ -78,22 +78,22 @@ client = InferenceClient("google/gemma-3-27b-it")
 
 def respond(message, history):
     # Step 1: Embed the user's question
-    #message_embedding = model.encode(message, convert_to_tensor=True)
+    message_embedding = model.encode(message, convert_to_tensor=True)
 
     # Step 2: Calculate similarity with knowledge chunks
-   # scores = util.cos_sim(message_embedding, chunk_embeddings)[0]
-    #top_k = 3  # You can adjust how many chunks you want to include
-   # top_results = torch.topk(scores, k=top_k)
+    scores = util.cos_sim(message_embedding, chunk_embeddings)[0]
+    top_k = 3  # You can adjust how many chunks you want to include
+    top_results = torch.topk(scores, k=top_k)
 
     # Step 3: Retrieve the top relevant knowledge chunks
     retrieved_knowledge = "\n".join([chunks[i] for i in top_results.indices])
 
-    # Step 4: Build system message with retrieved knowledge
+     Step 4: Build system message with retrieved knowledge
     system_message = (
         "You are a helpful chatbot named Scooby, kind of like the cartoon character but not too much.
         You know a lot about pets and their diets, and you only answer questions about pets.
         Use the following relevant knowledge to help answer the user's question"
-       # + retrieved_knowledge
+        + retrieved_knowledge
     )
 
     # Step 5: Compose message list for the LLM